Sort multiple tables inside Markdown file with text interspersed between them - bash

Question

There is a Markdown file with headings, text, and unsorted tables. I want to programmatically sort each table by ID, which is the 3rd column, in descending order, preferably using PowerShell or Bash. The table would remain in its place in the file.
# Heading
Text
| Col A | Col B | ID |
|---------|---------|----|
| Item 1A | Item 1B | 8 |
| Item 2A | Item 2B | 9 |
| Item 3A | Item 3B | 6 |
# Heading
Text
| Col A | Col B | ID |
|---------|---------|----|
| Item 4A | Item 4B | 3 |
| Item 5A | Item 5B | 2 |
| Item 6A | Item 6B | 4 |
I have no control over how the Markdown file is generated. Truly.
Ideally the file would remain in Markdown after the sort for additional processing. However, I explored these options without success:
Convert to JSON and sort (the solutions I tried didn't agree with tables)
Convert to HTML and sort (only found JavaScript solutions)
This script alone, while helpful, would need to be modified to parse through the Markdown file (having trouble finding understandable guidance on how to run a script on content between two strings)
The reason for command line (and not JavaScript on the HTML, for example) is that this transformation will take place in an Azure Release Pipeline. It is possible to add an Azure Function to the pipeline, which would allow me to run JavaScript code in the cloud, and I will pursue that if all else fails. I want to exhaust command-line options first because I am not very familiar with JavaScript or how to pass content between Functions and releases.
Thank you for any ideas.

Answer 1

By modifying the referred script, how about:
flush() {
printf "%s\n" "${lines[#]:0:2}"
printf "%s\n" "${lines[#]:2}" | sort -t \| -nr -k 4
lines=()
}
while IFS= read -r line; do
if [[ ${line:0:1} = "|" ]]; then
lines+=("$line")
else
(( ${#lines[#]} > 0 )) && flush
echo "$line"
fi
done < input.md
(( ${#lines[#]} > 0 )) && flush
Output:
# Heading
Text
| Col A | Col B | ID |
|---------|---------|----|
| Item 2A | Item 2B | 9 |
| Item 1A | Item 1B | 8 |
| Item 3A | Item 3B | 6 |
# Heading
Text
| Col A | Col B | ID |
|---------|---------|----|
| Item 6A | Item 6B | 4 |
| Item 4A | Item 4B | 3 |
| Item 5A | Item 5B | 2 |
BTW, if Perl is your option, here is an alternative:
perl -ne '
sub flush {
print splice(#ary, 0, 2); # print header lines
# sort the table with keying the ID by Schwartzian transform
print map { $_->[0] }
sort { $b->[1] <=> $a->[1] }
map { [$_, (split(/\s*\|\s*/))[3] ] }
#ary;
#ary = ();
}
# main loop
if (/^\|/) { # table section
push(#ary, $_);
} else { # other section
if ($#ary > 0) {
&flush;
} else {
print;
}
}
END {
if ($#ary > 0) { &flush; }
}
' input.md
Hope this helps.

Answer 2

If possible to identify markdown tables, a small 'awk' (or bash/python/perl) can filter the output. It assume each table has 2 header line.
awk -v 'FS="|" '
function cmp_id(i1, v1, i2, v2) {
return v1-v2 ;
}
function show () {
asorti(k, d, "cmp_id")
# for (i=1 ; i<=n; i++ ) print i, k[i], d[i] ;
# Print first 2 original header row, followed by sorted data lines
print s[1] ; print s[2]
for (i=1 ; i<=n; i++ ) if ( d[i]>=3 ) print s[d[i]] ;
n = 0
}
# Capture tables
/^\|/ { s[++n] = $0 ; k[n] = $4 ; next }
n > 0 { show() ; }
{ print }
END { show() ; }
'